| Abstract [eng] |
In modern data analysis are used many non-parametric methods for multivariate regression dependence of random variables for statistical analysis. Most popular method is the kernel estimator. Also quite popular are histospline, generalized additive models or local regression algorithms. For most popular non-parametric procedures for assessment in practice there is a problem of their choise. In this work, Monte Carlo approximation was used to perform various non-parametric estimates the accuracy of comparative analysis, in the case, when the valuation model is non-linear and describes the complex surface. |
